542_Final
Brittany Dymond, Hailey Fagerness, Cecilia Liang, Kathy Wang
3/16/2020
Fossil Fuels & Economic Development
Rationale of our two questions
How do fossil fuels relate to social and economic development in different countries? * Can we identify groups of countries with similar oil production and GDP and if/how oil production impacts a country’s GDP? * Can we identify groups of countries with similar population and fossil fuel usage and if/how population size affects fossil fuel usage?
Data: * Population * GDP per Capita * Oil Production (92 countries) * Fossil fuel use (as % of total electricity generating capacity)
QUESTION 1 CLUSTERING CODE START
RESEARCH QUESTION: ‘Can we identify groups of countries with similar oil production and GDP and if/how oil production impacts a country’s GDP?’
Data used: * Oil Production: [from U.S. Energy Information Administration] For calendar year 2019, on a comparable best-estimate basis * GDP per Capita: [from Wikipedia] Converted at market exchange rates to current U.S. dollars, divided by the population for the same year
Prep to cluster OilProduction and GDP_pc Getting data from github and initializing：
##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionRemoving rows where OilProduction == 0：
Clustering Part
Preparing to cluster oil production & GDP：
## OilProduction GDP_pc
## Albania 22915 5372
## Algeria 1348361 3980
## Angola 1769615 3037
## Argentina 510560 9887
## Australia 289749 53825
## Austria 15161 50022This is for replicability of results.
Partitioning Technique: PAM
Apply function and indicate the amount of clusters required
Clustering results
TABLE OF CLUSTERS：
##
## 1 2 3 4
## 58 14 17 3- Evaluate Results
AVG SILHOUETTES：
## Loading required package: ggplot2## Welcome! Want to learn more? See two factoextra-related books at https://goo.gl/ve3WBa## cluster size ave.sil.width
## 1 1 58 0.58
## 2 2 14 0.41
## 3 3 17 0.52
## 4 4 3 0.30
DECTECTING ANOMALIES：
## cluster neighbor sil_width
## Vietnam 1 3 0.7014283
## Congo, Republic of the 1 3 0.7006665
## Papua New Guinea 1 3 0.6977134
## Ghana 1 3 0.6972409
## Timor-Leste 1 3 0.6957549
## Tunisia 1 3 0.6955229Requesting negative silhouettes：
## cluster neighbor sil_width
## Italy 2 3 -0.277837863
## Romania 3 1 -0.008107932Hierarchizing/Agglomerative Technique: AGNES
Apply function and indicate the amount of clusters required
Clustering results
TABLE OF CLUSTERS:
##
## 1 2 3 4
## 59 14 16 3- Evaluate results
DENDOGRAM:
AVG SILHOUETTES:
## cluster size ave.sil.width
## 1 1 59 0.57
## 2 2 14 0.35
## 3 3 16 0.60
## 4 4 3 0.30
DECTECTING ANOMALIES:
## cluster neighbor sil_width
## Vietnam 1 3 0.6977023
## Congo, Republic of the 1 3 0.6970225
## Egypt 1 3 0.6946234
## Ghana 1 3 0.6884544
## Papua New Guinea 1 3 0.6880476
## Timor-Leste 1 3 0.6860947Requesting negative silhouettes:
## cluster neighbor sil_width
## Romania 1 3 -0.02769694
## Kuwait 2 3 -0.25164833Hierarchizing/Divisive Technique: DIANA
Apply function and indicate the amount of clusters required
Clustering results
TABLE OF CLUSTERS:
##
## 1 2 3 4
## 72 17 2 1- Evaluate results
DENDOGRAM:
AVG SILHOUETTES:
## cluster size ave.sil.width
## 1 1 72 0.72
## 2 2 17 0.50
## 3 3 2 0.70
## 4 4 1 0.00
DECTECTING ANOMALIES:
## cluster neighbor sil_width
## Mongolia 1 2 0.8295996
## Bolivia 1 2 0.8294036
## Ukraine 1 2 0.8291026
## Guatemala 1 2 0.8285038
## Tunisia 1 2 0.8284611
## Georgia 1 2 0.8284214Requesting negative silhouettes:
## [1] cluster neighbor sil_width
## <0 rows> (or 0-length row.names)Density Based Clustering: DBSCAN
HOW MANY OUTLIERS? (0 identified outliers)
## DBSCAN clustering for 92 objects.
## Parameters: eps = 0.03, minPts = 4
## The clustering contains 3 cluster(s) and 20 noise points.
##
## 0 1 2 3
## 20 53 14 5
##
## Available fields: cluster, eps, minPtsSave coordinates to original data frame:
Plot PAM：
Plot AGNES：
Plot DIANA：
Compare results visually：
## Loading required package: magrittr
- Plot DBSCAN：
Annotating:
Annotating Outliers:
BASED ON CLUSTERING, WE WILL USE DBSCAN. This cluster had high production &/OR high GDP (outliers).
## [1] "Australia" "Austria" "Brazil"
## [4] "Canada" "China" "Denmark"
## [7] "Iran" "Iraq" "Italy"
## [10] "Kazakhstan" "Kuwait" "Mexico"
## [13] "Netherlands" "Norway" "Oman"
## [16] "Russia" "Saudi Arabia" "United Arab Emirates"
## [19] "United Kingdom" "United States"This cluster had higher production & lower GDP.
## [1] "Albania"
## [2] "Algeria"
## [3] "Angola"
## [4] "Argentina"
## [5] "Azerbaijan"
## [6] "Bangladesh"
## [7] "Belarus"
## [8] "Belize"
## [9] "Bolivia"
## [10] "Bulgaria"
## [11] "Burma"
## [12] "Cameroon"
## [13] "Chad"
## [14] "Colombia"
## [15] "Congo, Democratic Republic of the"
## [16] "Congo, Republic of the"
## [17] "Ecuador"
## [18] "Egypt"
## [19] "Equatorial Guinea"
## [20] "Gabon"
## [21] "Georgia"
## [22] "Ghana"
## [23] "Guatemala"
## [24] "India"
## [25] "Indonesia"
## [26] "Kyrgyzstan"
## [27] "Libya"
## [28] "Malaysia"
## [29] "Mauritania"
## [30] "Mongolia"
## [31] "Morocco"
## [32] "Niger"
## [33] "Nigeria"
## [34] "Pakistan"
## [35] "Papua New Guinea"
## [36] "Peru"
## [37] "Philippines"
## [38] "Romania"
## [39] "Serbia"
## [40] "South Africa"
## [41] "Sudan"
## [42] "Suriname"
## [43] "Tajikistan"
## [44] "Thailand"
## [45] "Timor-Leste"
## [46] "Tunisia"
## [47] "Turkey"
## [48] "Turkmenistan"
## [49] "Ukraine"
## [50] "Uzbekistan"
## [51] "Venezuela"
## [52] "Vietnam"
## [53] "Yemen"This cluster had lower production & lower GDP.
## [1] "Bahrain" "Barbados" "Brunei"
## [4] "Chile" "Croatia" "Czechia"
## [7] "Greece" "Hungary" "Lithuania"
## [10] "Poland" "Slovakia" "Spain"
## [13] "Taiwan" "Trinidad and Tobago"This cluster had lower production & higher GDP.
## [1] "France" "Germany" "Israel" "Japan" "New Zealand"QUESTION 1 REGRESSION START
Hypothesis:
Model 1: GDP Per Capita ~ Oil Production
Model 2: GDP Per Capita ~ Oil Production + Continent
- Continuous Outcome – – GDP Per Capita
- Independent variable – – Oil Production
- Control variable – – Continent
- Rationale for hypothesis
- Oil infrastructure supports GDP
- OPEC // many economies heavily rely on oil income
- Oil price wars (like now with Saudi Arabia and Russia) impact oil prices and thus GDP
Preparing to regress Oil Production & GDP
## 'data.frame': 92 obs. of 12 variables:
## $ Country : chr "Albania" "Algeria" "Angola" "Arge"..
## $ fossilFuel_PctTotalElec: num 0.05 0.96 0.34 0.69 0.72 0.25 0.84 ..
## $ OilProduction : num 22915 1348361 1769615 510560 289749..
## $ Population : int 2880917 43053054 31825295 44780677 ..
## $ GDP_pc : int 5372 3980 3037 9887 53825 50022 468..
## $ Continent : Factor w/ 7 levels "Africa","Asia",..: 3..
## $ pam : Factor w/ 4 levels "1","2","3","4": 1 1 ..
## $ agn : Factor w/ 4 levels "1","2","3","4": 1 1 ..
## $ dia : Factor w/ 4 levels "1","2","3","4": 1 1 ..
## $ db : Factor w/ 4 levels "0","1","2","3": 2 2 ..
## $ dim1 : num -0.0795 -0.0701 -0.0722 -0.0399 0.2..
## $ dim2 : num 0.0111 -0.0435 -0.063 -0.0116 0.159..EXPLANATORY APPROACH
1.State the hypotheses
hypo1=formula(GDP_pc ~ OilProduction)
hypo2=formula(GDP_pc ~ OilProduction + Continent)2.Save colums needed and varify data types
## 'data.frame': 92 obs. of 3 variables:
## $ OilProduction: num 22915 1348361 1769615 510560 289749 ...
## $ GDP_pc : int 5372 3980 3037 9887 53825 50022 4689 25273 1905 18069 ...
## $ Continent : Factor w/ 7 levels "Africa","Asia",..: 3 1 1 7 6 3 4 2 2 5 ...3.Compute regression models
4.Hypothesis results
- First Hypothesis:
##
## Call:
## glm(formula = hypo1, family = "gaussian", data = DataRegGauss)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -20272 -11232 -7177 5968 61852
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.337e+04 1.882e+03 7.103 2.76e-10 ***
## OilProduction 1.673e-03 7.318e-04 2.286 0.0246 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for gaussian family taken to be 280755365)
##
## Null deviance: 2.6736e+10 on 91 degrees of freedom
## Residual deviance: 2.5268e+10 on 90 degrees of freedom
## AIC: 2054.7
##
## Number of Fisher Scoring iterations: 2- Second Hypothesis:
summary(gauss2)##
## Call:
## glm(formula = hypo2, family = "gaussian", data = DataRegGauss)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -29507 -8171 -1380 5466 47217
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.227e+03 3.288e+03 0.677 0.500143
## OilProduction 2.140e-03 6.498e-04 3.294 0.001448 **
## ContinentAsia 7.450e+03 4.231e+03 1.761 0.081897 .
## ContinentEurope 2.500e+04 4.421e+03 5.656 2.09e-07 ***
## ContinentEurope/Asia -2.640e+02 7.171e+03 -0.037 0.970725
## ContinentNorth America 1.499e+04 6.409e+03 2.340 0.021675 *
## ContinentOceania 2.990e+04 8.669e+03 3.449 0.000882 ***
## ContinentSouth America 3.519e+03 5.678e+03 0.620 0.537102
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for gaussian family taken to be 193147596)
##
## Null deviance: 2.6736e+10 on 91 degrees of freedom
## Residual deviance: 1.6224e+10 on 84 degrees of freedom
## AIC: 2026
##
## Number of Fisher Scoring iterations: 25.Searching for a better model
## Analysis of Deviance Table
##
## Model 1: GDP_pc ~ OilProduction
## Model 2: GDP_pc ~ OilProduction + Continent
## Resid. Df Resid. Dev Df Deviance Pr(>Chi)
## 1 90 2.5268e+10
## 2 84 1.6224e+10 6 9043584826 2.03e-08 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Model for the Second hypothesis is chosen. This is the RSquared:
## [1] 0.34258156.Verify the situation of chosen model:
6.1. Linearity between dependent variable and predictors is assumed, then these dots should follow a linear and horizontal trend:
The linear trend is not obvious, and the distribution range goes wider when the predicted values increase. I’d like to say it represents the linearity between our variables in a certain level. Further research upon outliers are necessary.
6.2. Normality of residuals is assumed:
Visual exploration
Mathematical exploration:
##
## Shapiro-Wilk normality test
##
## data: gauss2$residuals
## W = 0.94464, p-value = 0.0006816.3. Homoscedasticity is assumed, so check if residuals are spread equally along the ranges of predictors
Visual exploration:
Mathematical exploration:
## Loading required package: zoo##
## Attaching package: 'zoo'## The following objects are masked from 'package:base':
##
## as.Date, as.Date.numeric##
## studentized Breusch-Pagan test
##
## data: gauss2
## BP = 19.735, df = 7, p-value = 0.0061716.4. We assume that there is no colinearity, that is, that the predictors are not correlated.
## Loading required package: carData##
## Attaching package: 'car'## The following object is masked from 'package:dplyr':
##
## recode## GVIF Df GVIF^(1/(2*Df))
## OilProduction 1.146212 1 1.070613
## Continent 1.146212 6 1.0114376.5. Analize the effect of atypical values. Determine if outliers (points that are far from the rest, but still in the trend) or high-leverage points (far from the trend but close to the rest) are influential
Visual exploration:
Querying:
gaussInf=as.data.frame(influence.measures(gauss2)$is.inf)
gaussInf[gaussInf$cook.d,]## [1] dfb.1_ dfb.OlPr dfb.CntA dfb.CntE dfb.CE/A dfb.CnNA dfb.CntO
## [8] dfb.CnSA dffit cov.r cook.d hat
## <0 rows> (or 0-length row.names)- Finally, a nice summary plot of our work
## Registered S3 methods overwritten by 'lme4':
## method from
## cooks.distance.influence.merMod car
## influence.merMod car
## dfbeta.influence.merMod car
## dfbetas.influence.merMod car## Learn more about sjPlot with 'browseVignettes("sjPlot")'.
PREDICTIVE APPROACH
- Splitting the data set
## Loading required package: lattice- Regress with train data
##
## Call:
## NULL
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -29470 -9388 -1480 7635 44909
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.204e+03 4.180e+03 0.527 0.59982
## OilProduction 2.670e-03 8.371e-04 3.189 0.00221 **
## ContinentAsia 8.393e+03 5.127e+03 1.637 0.10657
## ContinentEurope 2.646e+04 5.685e+03 4.655 1.68e-05 ***
## `ContinentEurope/Asia` -1.646e+03 8.083e+03 -0.204 0.83925
## `ContinentNorth America` 1.532e+04 7.708e+03 1.988 0.05109 .
## ContinentOceania 2.986e+04 9.604e+03 3.109 0.00280 **
## `ContinentSouth America` 3.651e+03 7.890e+03 0.463 0.64509
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for gaussian family taken to be 224537488)
##
## Null deviance: 2.3585e+10 on 71 degrees of freedom
## Residual deviance: 1.4370e+10 on 64 degrees of freedom
## AIC: 1598.4
##
## Number of Fisher Scoring iterations: 2- Evaluate performance
## RMSE Rsquared MAE
## 10322.575720 0.536489 7547.290591QUESTION 2 CLUSTERING START
RESEARCH QUESTION: ‘Can we identify groups of countries with similar population and fossil fuel usage and if/how population size affects fossil fuel usage?’
Data used: * fossilFuel_PctTotalElec: [from CIA World Factbook] percentage of total electricity generating capacity that comes from fossil fuels * Population: [UN Dept of Economic and Social Affairs] World population estimates
Prep to cluster fossilFuel_PctTotalElec and Population
## fossilFuel_PctTotalElec Population
## Albania 0.05 2880917
## Algeria 0.96 43053054
## Angola 0.34 31825295
## Argentina 0.69 44780677
## Australia 0.72 25203198
## Austria 0.25 8955102Set random seed for replicability of results:
Setting distance matrix:
Defining number of clusters for each method (NumCluster = 5) Clustering via pam method:
Adding pam results to original DF (DFnew1)
REPORT: Table of Cluster:
##
## 1 2 3 4 5
## 14 26 30 20 2REPORT: Evaluate Results:
## cluster size ave.sil.width
## 1 1 14 0.38
## 2 2 26 0.63
## 3 3 30 0.36
## 4 4 20 0.57
## 5 5 2 0.86
REPORT: Detecting Anomalies
Saving individual silhouettes
## cluster neighbor sil_width
## Tajikistan 1 4 0.6090716
## Albania 1 4 0.6060691
## Norway 1 4 0.5978116
## Timor-Leste 1 4 0.5667970
## Congo, Democratic Republic of the 1 4 0.5525901
## France 1 4 0.5468275Requesting negative silhouettes:
## cluster neighbor sil_width
## Angola 1 4 -0.2800818
## Georgia 1 4 -0.3377824
## Ghana 3 4 -0.1223473Cluster via agnes method; indicate number of clusters (NumCluster):
Adding agn results to original DF (DFnew1)
REPORT: Table of clusters:
##
## 1 2 3 4 5
## 12 19 21 38 2Evaluating results:
REPORT: Average silhouettes
## cluster size ave.sil.width
## 1 1 12 0.47
## 2 2 19 0.79
## 3 3 21 0.60
## 4 4 38 0.23
## 5 5 2 0.87
REPORT: Detecting anomalies
## cluster neighbor sil_width
## Tajikistan 1 3 0.6486823
## Albania 1 3 0.6461141
## Norway 1 3 0.6392016
## Timor-Leste 1 3 0.6056001
## Congo, Democratic Republic of the 1 3 0.5963339
## France 1 3 0.5571241Requesting negative silhouettes:
## cluster neighbor sil_width
## Colombia 1 3 -0.03279474
## Indonesia 4 2 -0.01212078
## Iran 4 2 -0.03057378
## Chile 4 3 -0.06866262
## Azerbaijan 4 2 -0.14697731
## South Africa 4 2 -0.15226216
## Ghana 4 3 -0.17299110
## Greece 4 3 -0.27237486
## Uzbekistan 4 2 -0.28525827
## Kazakhstan 4 2 -0.31159271
## Mongolia 4 2 -0.41514761Cluster via diana method; indicate number of clusters (NumCluster):
Adding diana results to original DF (DFnew1):
REPORT: Table of clusters
##
## 1 2 3 4 5
## 12 26 23 29 2REPORT: Dendrogram 
REPORT: Average silhouettes
## cluster size ave.sil.width
## 1 1 12 0.49
## 2 2 26 0.62
## 3 3 23 0.52
## 4 4 29 0.39
## 5 5 2 0.86
REPORT: Detecting anomalies:
## cluster neighbor sil_width
## Tajikistan 1 3 0.6564827
## Albania 1 3 0.6539019
## Norway 1 3 0.6467797
## Timor-Leste 1 3 0.6133818
## Congo, Democratic Republic of the 1 3 0.6057145
## France 1 3 0.5720875Requesting negative silhouettes:
## [1] cluster neighbor sil_width
## <0 rows> (or 0-length row.names)Cluster via DBSCAN method; indicate minimum neighbors (4):
Setting distance (epsilon):
REPORT: Number of clusters and outliers produced
## DBSCAN clustering for 92 objects.
## Parameters: eps = 0.03, minPts = 4
## The clustering contains 4 cluster(s) and 11 noise points.
##
## 0 1 2 3 4
## 11 4 52 4 21
##
## Available fields: cluster, eps, minPtsSaving results:
Comparing clusters
Prepare a bidimensional map:
View bidimensional map:
Results from pam:
Results from agnes:
Results from diana：
Compare visually:
Viewing pam, agnes, and diana plots side by side
Plot results from DBSCAN：
Annotating graph with country names：
Annotating just the outlier countries：
CHOOSING DIANA METHOD DUE TO HAVING ZERO NEGATIVE SILHOUETTES。
QUESTION 2 REGRESSION CODE START
- Hypothesis:
- Model 1: FF ~ Population
- Model 2: FF ~ Population + Developed
- Method:
- Binary Outcome – – FF usage (Median percentage FF use of total electricity capacity)
- Control variable – – Developed (Median GDP per capita)
- Independent variable – – Population
- Rationale for hypothesis:
- Larger populations would exhibit higher fossil fuel usage as a percent of total electricity capacity
Changing dtype for population：
## Country fossilFuel_PctTotalElec OilProduction
## "character" "numeric" "numeric"
## Population GDP_pc
## "numeric" "integer"Changing dtype for GDP_pc(gdp):
## Country fossilFuel_PctTotalElec OilProduction
## "character" "numeric" "numeric"
## Population GDP_pc
## "numeric" "numeric"Filtering out non-oil producing countries & creating new DF (teamnew):
## Country fossilFuel_PctTotalElec
## 2 Albania 0.05
## 3 Algeria 0.96
## 4 Angola 0.34
## 6 Argentina 0.69
## 9 Australia 0.72
## 10 Austria 0.25
## 11 Azerbaijan 0.84
## 13 Bahrain 1.00
## 14 Bangladesh 0.97
## 15 Barbados 0.93
## 16 Belarus 0.96
## 18 Belize 0.51
## 21 Bolivia 0.76
## 24 Brazil 0.17
## 25 Brunei 1.00
## 26 Bulgaria 0.39
## 28 Burma 0.39
## 32 Cameroon 0.52
## 33 Canada 0.23
## 35 Chad 0.98
## 36 Chile 0.59
## 37 China 0.62
## 38 Colombia 0.29
## 40 Congo, Democratic Republic of the 0.02
## 41 Congo, Republic of the 0.64
## 43 Croatia 0.45
## 45 Czechia 0.60
## 46 Denmark 0.46
## 50 Ecuador 0.43
## 51 Egypt 0.91
## 53 Equatorial Guinea 0.61
## 60 France 0.17
## 61 Gabon 0.51
## 63 Georgia 0.35
## 64 Germany 0.41
## 65 Ghana 0.58
## 66 Greece 0.57
## 68 Guatemala 0.41
## 75 Hungary 0.64
## 77 India 0.71
## 78 Indonesia 0.85
## 79 Iran 0.84
## 80 Iraq 0.91
## 82 Israel 0.95
## 83 Italy 0.54
## 85 Japan 0.71
## 87 Kazakhstan 0.86
## 91 Kuwait 1.00
## 92 Kyrgyzstan 0.24
## 98 Libya 1.00
## 99 Lithuania 0.73
## 104 Malaysia 0.78
## 109 Mauritania 0.65
## 111 Mexico 0.71
## 114 Mongolia 0.87
## 116 Morocco 0.68
## 121 Netherlands 0.75
## 122 New Zealand 0.23
## 124 Niger 0.95
## 125 Nigeria 0.80
## 127 Norway 0.03
## 128 Oman 1.00
## 129 Pakistan 0.62
## 131 Papua New Guinea 0.63
## 133 Peru 0.61
## 134 Philippines 0.67
## 135 Poland 0.79
## 139 Romania 0.47
## 140 Russia 0.68
## 147 Saudi Arabia 1.00
## 149 Serbia 0.65
## 153 Slovakia 0.36
## 156 South Africa 0.85
## 158 Spain 0.47
## 160 Sudan 0.44
## 161 Suriname 0.61
## 164 Taiwan 0.79
## 165 Tajikistan 0.06
## 167 Thailand 0.76
## 168 Timor-Leste 0.00
## 171 Trinidad and Tobago 1.00
## 172 Tunisia 0.94
## 173 Turkey 0.53
## 174 Turkmenistan 1.00
## 177 Ukraine 0.65
## 178 United Arab Emirates 0.99
## 179 United Kingdom 0.50
## 180 United States 0.70
## 182 Uzbekistan 0.86
## 184 Venezuela 0.51
## 185 Vietnam 0.56
## 186 Yemen 0.79
## OilProduction Population GDP_pc
## 2 22915 2880917 5372
## 3 1348361 43053054 3980
## 4 1769615 31825295 3037
## 6 510560 44780677 9887
## 9 289749 25203198 53825
## 10 15161 8955102 50022
## 11 833538 10047718 4689
## 13 40000 1641172 25273
## 14 4189 163046161 1905
## 15 1000 287025 18069
## 16 25000 9452411 6603
## 18 2000 390353 4925
## 21 58077 11513100 3670
## 24 2515459 211049527 8796
## 25 109117 433285 27871
## 26 1000 7000119 9518
## 28 15000 54045420 1244
## 32 93205 25876380 1514
## 33 3662694 37411047 46212
## 35 110156 15946876 861
## 36 4423 18952038 15399
## 37 3980650 1433783686 10098
## 38 897784 50339443 6508
## 40 20000 86790567 500
## 41 308363 5380508 2534
## 43 13582 4130304 14949
## 45 2333 10689209 23213
## 46 140637 5771876 59795
## 50 548421 17373662 6249
## 51 490000 100388073 3046
## 53 227000 1355986 8927
## 60 16418 65129728 41760
## 61 210820 2172579 8112
## 63 400 3996765 4289
## 64 46839 83517045 46563
## 65 100549 28833629 2223
## 66 3172 10473455 19974
## 68 8977 17581472 4616
## 75 13833 9684679 17463
## 77 715459 1366417754 2171
## 78 833667 270625568 4163
## 79 3990956 82913906 5506
## 80 4451516 39309783 5738
## 82 390 8519377 42823
## 83 70675 60550075 32946
## 85 3918 126860301 40846
## 87 1595199 18551427 9139
## 91 2923825 4207083 29266
## 92 1000 6415850 1292
## 98 1003000 6777452 5019
## 99 2000 2759627 19266
## 104 661240 31949777 11136
## 109 5000 4525696 1392
## 111 2186877 127575529 10118
## 114 23426 3225167 4132
## 116 160 36471769 3345
## 121 18087 17097130 52367
## 122 35574 4783063 40634
## 124 13000 23310715 405
## 125 1999885 200963599 2222
## 127 1647975 5378857 77975
## 128 1006841 4974986 17791
## 129 80000 216565318 1388
## 131 56667 8776109 2742
## 133 40266 32510453 7046
## 134 20000 108116615 3294
## 135 20104 37887768 14901
## 139 504000 19364557 12482
## 140 10800000 145872256 11162
## 147 12000000 34268528 22865
## 149 20000 8772235 7397
## 153 200 5457013 19547
## 156 2000 58558270 6100
## 158 2667 46736776 29961
## 160 255000 42813238 714
## 161 17000 581372 6310
## 164 196 23773876 24827
## 165 180 9321018 877
## 167 257525 69037513 7791
## 168 60661 1293119 2262
## 171 60090 1394973 16365
## 172 48757 11694719 3287
## 173 49497 83429615 8957
## 174 230779 5942089 7816
## 177 31989 43993638 3592
## 178 3106077 9770529 37749
## 179 939760 67530172 41030
## 180 15043000 329064917 65111
## 182 52913 32981716 1831
## 184 2276967 28515829 2547
## 185 301850 96462106 2740
## 186 22000 29161922 943Converting USDollar to a factor variable
Calling new variable ‘Developed’
Converting fossilFuel_PctTotalElec to a factor variable
Calling new variable ‘FF’
Checking dtypes:
## 'data.frame': 92 obs. of 7 variables:
## $ Country : chr "Albania" "Algeria" "Angola" "Arge"..
## $ fossilFuel_PctTotalElec: num 0.05 0.96 0.34 0.69 0.72 0.25 0.84 ..
## $ OilProduction : num 22915 1348361 1769615 510560 289749..
## $ Population : num 2880917 43053054 31825295 44780677 ..
## $ GDP_pc : num 5372 3980 3037 9887 53825 ...
## $ Developed : Factor w/ 2 levels "0","1": 1 1 1 2 2 2 ..
## $ FF : Factor w/ 2 levels "0","1": 1 2 1 2 2 1 ..Defining ‘Population’ as independent variable:
Defining columns needed：
Verify dtypes for colsNeededDico:
## 'data.frame': 92 obs. of 3 variables:
## $ FF : Factor w/ 2 levels "0","1": 1 2 1 2 2 1 2 2 2 2 ...
## $ Population: num 2880917 43053054 31825295 44780677 25203198 ...
## $ Developed : Factor w/ 2 levels "0","1": 1 1 1 2 2 2 1 2 1 2 ...Create subset
Rename indexes by country
Define & compute regression models
Results of hypo3:
At p-value of 0.634, this model is not statistically significant
##
## Call:
## glm(formula = hypo3, family = "binomial", data = DataRegLogis)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.436 -1.129 -1.126 1.223 1.230
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.231e-01 2.217e-01 -0.555 0.579
## Population 4.972e-10 1.045e-09 0.476 0.634
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 127.37 on 91 degrees of freedom
## Residual deviance: 127.13 on 90 degrees of freedom
## AIC: 131.13
##
## Number of Fisher Scoring iterations: 4Results of hypo4:
At p-values of 0.631 and 0.673, this model also is not statistically significant:
##
## Call:
## glm(formula = hypo4, family = "binomial", data = DataRegLogis)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.479 -1.141 -1.089 1.192 1.268
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.120e-01 3.063e-01 -0.692 0.489
## Population 5.028e-10 1.048e-09 0.480 0.631
## Developed1 1.768e-01 4.184e-01 0.422 0.673
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 127.37 on 91 degrees of freedom
## Residual deviance: 126.95 on 89 degrees of freedom
## AIC: 132.95
##
## Number of Fisher Scoring iterations: 4Analysis of variance between models:
## Analysis of Deviance Table
##
## Model 1: FF ~ Population
## Model 2: FF ~ Population + Developed
## Resid. Df Resid. Dev Df Deviance Pr(>Chi)
## 1 90 127.13
## 2 89 126.95 1 0.17869 0.6725Recommendations
First Question
Oil production could be an important component of GDP, but higher oil production rate does not lead to higher GDP. If we want to evaluate the relationship between GDP and oil production, we also need to know what is the percentage of the GDP generated by oil production. Same level variables are more easy to be compared Too many countries that their oil production is close to zero Try other control variables like export/import Higher oil production does not lead to higher GDP necessarily
Second Question
Neither model is statistically significant; no further analysis required.
Recommendations for future analysis of question #2 include: Incorporate country-specific income levels as an additional variable Remove major outliers from sample population *Use actual fossil fuel usage data in lieu of ratios